pith. sign in

arxiv: 2508.19177 · v2 · submitted 2025-08-26 · 🧮 math.NA · cs.NA

Stoch-IDENT: New Method and Mathematical Analysis for Identifying SPDEs from Data

Jianbo Cui , Roy Y. He This is my paper

Pith reviewed 2026-05-18 20:40 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords SPDE identificationstochastic partial differential equationssparse regressionquadratic measurementsQuadratic Subspace Pursuitidentifiability analysistrajectory dataWiener processes
0
0 comments X

The pith

Stoch-IDENT recovers both drift and diffusion coefficients of high-order linear and nonlinear SPDEs from trajectory data by solving a quadratic sparse regression problem.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a framework called Stoch-IDENT that identifies the full structure of stochastic partial differential equations, including the deterministic drift and the stochastic diffusion driven by time-dependent Wiener processes. It first generalizes identifiability results from deterministic PDEs by examining the spectral properties of the solution mean and covariance operators for linear constant-coefficient SPDEs and the dimension of the solution space for parabolic and hyperbolic types. The drift is recovered through a sample-mean extension of existing regression techniques, while the diffusion is isolated by forming quadratic measurements from drift residuals and feature covariances and then solving the resulting non-convex sparse regression via a new greedy algorithm, Quadratic Subspace Pursuit, whose stable support recovery is proved under suitable conditions. A sympathetic reader would care because the approach works for both additive and multiplicative noise and for high-order nonlinear equations, turning multiple observed trajectories into a complete stochastic model without requiring prior knowledge of the noise structure.

Core claim

The paper shows that SPDE identifiability from trajectory data follows from the spectral properties of the mean and covariance for linear cases with constant coefficients, generalizing deterministic PDE theory, and that the diffusion term can be recovered by casting the problem as sparse regression with quadratic measurements induced from drift residuals and feature covariances; the new Quadratic Subspace Pursuit algorithm solves this optimization and enjoys stable support recovery under certain conditions on the data.

What carries the argument

Quadratic Subspace Pursuit (QSP), a greedy algorithm that selects atoms to solve the non-convex, non-smooth sparse regression problem whose measurements are quadratic forms built from drift residuals and empirical covariances.

If this is right

  • The method accommodates both additive and multiplicative noise structures in high-order SPDEs.
  • QSP provides stable support recovery for the diffusion coefficients under conditions on the number and quality of trajectories.
  • The sample-mean approach for the drift term extends deterministic PDE identification while accounting for stochastic effects.
  • Numerical tests on various linear and nonlinear SPDEs confirm quantitative accuracy and qualitative fidelity of the recovered equations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same quadratic-measurement construction could be tested on inverse problems for stochastic ordinary differential equations or on data from real physical systems where only noisy trajectories are available.
  • If the identifiability analysis for linear constant-coefficient cases extends to variable coefficients, the framework might apply to a broader class of time-inhomogeneous SPDEs.
  • One could examine whether replacing the greedy QSP step with a convex relaxation changes the support-recovery guarantees or computational cost on the same quadratic measurements.

Load-bearing premise

The observed data must consist of sufficiently many independent trajectories so that sample means and empirical covariances converge to the true mean and covariance operators, allowing the quadratic measurements to isolate the diffusion coefficients without finite-sample bias.

What would settle it

Apply Stoch-IDENT to a large number of independent trajectories generated from a known linear SPDE whose drift and diffusion coefficients are given in advance; if the recovered diffusion support differs from the true support even though the number of trajectories is high and the theoretical conditions for QSP are met, the central recovery claim is false.

Figures

Figures reproduced from arXiv: 2508.19177 by Jianbo Cui, Roy Y. He.

Figure 1
Figure 1. Figure 1: Identification performance versus the number of trajectories. Column (a) Stochastic trans [PITH_FULL_IMAGE:figures/full_fig_p022_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Identification of the stochastic NLS equation ( [PITH_FULL_IMAGE:figures/full_fig_p023_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between a solution path of stochastic Allen-Cahn ( [PITH_FULL_IMAGE:figures/full_fig_p025_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison between identification using trajectories with or without random initializations. [PITH_FULL_IMAGE:figures/full_fig_p026_4.png] view at source ↗
read the original abstract

In this paper, we propose Stoch-IDENT, a novel framework for identifying stochastic partial differential equations (SPDEs) from observational data. Our method can handle linear and nonlinear high-order SPDEs driven by time-dependent Wiener processes, accommodating both additive and multiplicative noise structures. To investigate the identifiability of SPDEs from trajectory data, we analyze the spectral properties of the solution's mean and covariance for linear SPDEs with constant coefficients, as well as the dimension of the solution space for parabolic and hyperbolic types, generalizing the identifiability theory for deterministic PDEs. Algorithmically, the drift term is identified via a sample-mean generalization of existing methods for PDE identification. For the diffusion term, we formulate a sparse regression problem with quadratic measurements induced from drift residuals and feature covariances. To address this challenging non-convex and non-smooth optimization, we develop a new greedy algorithm, Quadratic Subspace Pursuit (QSP), and prove that QSP enjoys stable support recovery under certain conditions. We validate Stoch-IDENT on various SPDEs, demonstrating its effectiveness through quantitative and qualitative evaluations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes Stoch-IDENT, a framework for recovering both drift and diffusion terms of linear and nonlinear high-order SPDEs driven by time-dependent Wiener processes (additive or multiplicative noise) from trajectory data. Drift recovery generalizes deterministic sample-mean methods; diffusion is cast as a quadratic-measurement sparse regression solved by the new Quadratic Subspace Pursuit (QSP) algorithm, for which stable support recovery is proven under stated conditions. Identifiability is established via spectral properties of the mean and covariance operators together with solution-space dimension for linear constant-coefficient parabolic and hyperbolic SPDEs, generalizing deterministic PDE theory. Numerical experiments on several SPDEs are reported.

Significance. If the central claims hold, the work would advance data-driven discovery of SPDEs by supplying a unified algorithmic pipeline that handles both drift and diffusion, together with a new greedy solver (QSP) for quadratic measurements and a partial identifiability theory for the linear case. The explicit proof of stable support recovery for QSP and the generalization of deterministic identifiability arguments are concrete strengths that would be of interest to the numerical analysis and stochastic modeling communities.

major comments (3)
  1. [Identifiability analysis] Identifiability section: spectral and dimension arguments are supplied only for linear constant-coefficient SPDEs. No corresponding uniqueness, spectral, or solution-space analysis is given for the nonlinear SPDEs whose identification is claimed in the abstract and algorithmic sections; the central claim that the full framework identifies nonlinear SPDEs therefore rests on an unproven extension of the linear theory.
  2. [QSP algorithm and proof] QSP theorem: the conditions guaranteeing stable support recovery are referenced in the abstract but not restated or verified for the specific quadratic measurement operators constructed from drift residuals and empirical covariances in the SPDE examples; without this verification the recovery guarantee does not directly apply to the reported experiments.
  3. [Algorithmic framework] Drift-then-diffusion pipeline: the diffusion step uses residuals from the sample-mean drift estimate; no quantitative bound is provided on how finite-sample bias or variance in the drift step propagates into the quadratic measurements or the subsequent QSP recovery, which is load-bearing for the claimed robustness on finite trajectory data.
minor comments (2)
  1. [Notation and preliminaries] Notation for the quadratic measurement matrix and the precise definition of the feature library for nonlinear terms should be introduced earlier and used consistently across the algorithmic and theoretical sections.
  2. [Abstract] The abstract states that QSP enjoys stable support recovery 'under certain conditions'; these conditions should be summarized in one sentence in the abstract for immediate clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful reading of our manuscript and the constructive comments. We address each of the major comments below and indicate the revisions we plan to make.

read point-by-point responses

  1. Referee: Identifiability section: spectral and dimension arguments are supplied only for linear constant-coefficient SPDEs. No corresponding uniqueness, spectral, or solution-space analysis is given for the nonlinear SPDEs whose identification is claimed in the abstract and algorithmic sections; the central claim that the full framework identifies nonlinear SPDEs therefore rests on an unproven extension of the linear theory.

    Authors: We appreciate this observation. The identifiability analysis provided in the manuscript focuses on linear constant-coefficient SPDEs, extending the deterministic theory through spectral properties of the mean and covariance operators and the dimension of the solution space. For nonlinear SPDEs, the identification relies on the data-driven sparse regression framework, which is validated through numerical experiments on several nonlinear examples. To better align the claims with the provided analysis, we will revise the abstract and relevant sections to specify that the rigorous identifiability results apply to linear SPDEs, while the method is designed and tested for both linear and nonlinear cases. This revision will clarify the theoretical scope. revision: yes

  2. Referee: QSP theorem: the conditions guaranteeing stable support recovery are referenced in the abstract but not restated or verified for the specific quadratic measurement operators constructed from drift residuals and empirical covariances in the SPDE examples; without this verification the recovery guarantee does not directly apply to the reported experiments.

    Authors: The stable support recovery theorem for QSP is stated in general terms in the manuscript, applicable to quadratic measurement models satisfying the given conditions. In the context of SPDE identification, the quadratic measurements are derived from the residuals after drift estimation and the covariance structure of the features. We will include an additional discussion or appendix that verifies or discusses how the general conditions of the theorem are met for the operators used in the numerical experiments, thereby strengthening the connection between the theory and the reported results. revision: yes

  3. Referee: Drift-then-diffusion pipeline: the diffusion step uses residuals from the sample-mean drift estimate; no quantitative bound is provided on how finite-sample bias or variance in the drift step propagates into the quadratic measurements or the subsequent QSP recovery, which is load-bearing for the claimed robustness on finite trajectory data.

    Authors: We agree that a quantitative analysis of error propagation from the drift estimation to the diffusion recovery would provide stronger theoretical support for the robustness claims. Currently, the manuscript demonstrates robustness through numerical experiments with varying numbers of trajectories and noise intensities. Providing a full perturbation bound is a non-trivial extension that would involve analyzing the sensitivity of the quadratic measurements and the QSP algorithm to perturbations in the residuals. We will add a paragraph in the discussion section to acknowledge this limitation and to outline the conditions under which the method is expected to perform reliably based on the empirical evidence. revision: partial

Circularity Check

0 steps flagged

No circularity detected; derivation chain remains self-contained

full rationale

The paper grounds its identifiability claims for linear constant-coefficient SPDEs in an analysis of spectral properties of the mean and covariance operators together with solution-space dimension, explicitly generalizing prior deterministic PDE results rather than deriving them from the present method's outputs. The algorithmic pipeline separates drift recovery (via sample-mean extension of existing deterministic techniques) from diffusion recovery (via quadratic measurements on residuals and covariances, solved by the independently analyzed QSP algorithm whose support-recovery guarantee is stated under explicit conditions). No step reduces a claimed prediction or uniqueness result to a fitted parameter, self-citation chain, or definitional tautology; the nonlinear extension is presented as an algorithmic application without asserting that the linear spectral theory forces it by construction. The overall framework therefore contains independent mathematical content and does not collapse to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the assumption that the observed data are generated by an SPDE whose drift and diffusion operators belong to a known finite dictionary of candidate terms, and that the noise is driven by a Wiener process whose statistics allow the covariance operator to be estimated from trajectories.

axioms (1)
  • domain assumption The solution process admits well-defined mean and covariance operators whose spectral properties determine identifiability for linear constant-coefficient SPDEs.
    Invoked in the analysis of spectral properties and dimension of solution space for parabolic and hyperbolic types.

pith-pipeline@v0.9.0 · 5723 in / 1361 out tokens · 27270 ms · 2026-05-18T20:40:10.405981+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

60 extracted references · 60 canonical work pages

  1. [1]

    D. C. Antonopoulou, D. Farazakis, and G. Karali , Malliavin calculus for the stochastic Cahn-Hilliard/Allen-Cahn equation with unbounded noise diffusion , J. Differential Equations, 265 (2018), pp. 3168–3211

    work page 2018

  2. [2]

    Balan, B

    R. Balan, B. G. Bodmann, P. G. Casazza, and D. Edidin , Painless reconstruction from magnitudes of frame coefficients, Journal of Fourier Analysis and Applications, 15 (2009), pp. 488– 501. 36

    work page 2009

  3. [3]

    Balan, P

    R. Balan, P. Casazza, and D. Edidin , On signal reconstruction without phase , Applied and Computational Harmonic Analysis, 20 (2006), pp. 345–356

    work page 2006

  4. [4]

    A. S. Bandeira, E. Dobriban, D. G. Mixon, and W. F. Sawin , Certifying the restricted isometry property is hard , IEEE transactions on information theory, 59 (2013), pp. 3448–3450

    work page 2013

  5. [5]

    Beck and Y

    A. Beck and Y. C. Eldar , Sparsity constrained nonlinear optimization: Optimality conditions and algorithms, SIAM Journal on Optimization, 23 (2013), pp. 1480–1509

    work page 2013

  6. [6]

    Berg and K

    J. Berg and K. Nystr ¨om, Data-driven discovery of PDEs in complex datasets , Journal of Computational Physics, 384 (2019), pp. 239–252

    work page 2019

  7. [7]

    Bl ¨omker and A

    D. Bl ¨omker and A. Jentzen , Galerkin approximations for the stochastic Burgers equation , SIAM J. Numer. Anal., 51 (2013), pp. 694–715

    work page 2013

  8. [8]

    Boninsegna, F

    L. Boninsegna, F. N ¨uske, and C. Clementi , Sparse learning of stochastic dynamical equa- tions, The Journal of chemical physics, 148 (2018)

    work page 2018

  9. [9]

    Br´ehier, J

    C.-E. Br´ehier, J. Cui, and J. Hong, Strong convergence rates of semidiscrete splitting approx- imations for the stochastic Allen-Cahn equation , IMA J. Numer. Anal., 39 (2019), pp. 2096–2134

    work page 2019

  10. [10]

    S. L. Brunton, J. L. Proctor, and J. N. Kutz , Discovering governing equations from data by sparse identification of nonlinear dynamical systems , Proceedings of the national academy of sciences, 113 (2016), pp. 3932–3937

    work page 2016

  11. [11]

    E. J. Candes and T. Tao, Decoding by linear programming, IEEE transactions on information theory, 51 (2005), pp. 4203–4215

    work page 2005

  12. [12]

    J. Chen, M. K. Ng, and Z. Liu , Solving quadratic systems with full-rank matrices using sparse or generative priors , IEEE Transactions on Signal Processing, (2025)

    work page 2025

  13. [13]

    Cui and J

    J. Cui and J. Hong, Analysis of a splitting scheme for damped stochastic nonlinear Schr¨ odinger equation with multiplicative noise , SIAM J. Numer. Anal., 56 (2018), pp. 2045–2069

    work page 2018

  14. [14]

    J. Cui, J. Hong, and Z. Liu , Strong convergence rate of finite difference approximations for stochastic cubic Schr¨ odinger equations, J. Differential Equations, 263 (2017), pp. 3687–3713

    work page 2017

  15. [15]

    Cui and L

    J. Cui and L. Sun , Quantifying the effect of random dispersion for logarithmic Schr¨ odinger equation, SIAM/ASA J. Uncertain. Quantif., 12 (2024), pp. 579–613

    work page 2024

  16. [16]

    Da Prato and J

    G. Da Prato and J. Zabczyk, Stochastic equations in infinite dimensions , vol. 44 of Encyclo- pedia of Mathematics and its Applications, Cambridge University Press, Cambridge, 1992

    work page 1992

  17. [17]

    D’agostino and E

    R. D’agostino and E. S. Pearson , Tests for departure from normality. empirical results for the distributions of b2 and √b1, Biometrika, 60 (1973), pp. 613–622

    work page 1973

  18. [18]

    Dai and O

    W. Dai and O. Milenkovic , Subspace pursuit for compressive sensing signal reconstruction , IEEE transactions on Information Theory, 55 (2009), pp. 2230–2249

    work page 2009

  19. [19]

    Debussche and J

    A. Debussche and J. Printems , Numerical simulation of the stochastic Korteweg-de Vries equation, Phys. D, 134 (1999), pp. 200–226

    work page 1999

  20. [20]

    C. Dong, G. Xu, G. Han, B. J. Bethel, W. Xie, and S. Zhou , Recent developments in artificial intelligence in oceanography, Ocean-Land-Atmosphere Research, (2022)

    work page 2022

  21. [21]

    D. L. Donoho and M. Elad , Optimally sparse representation in general (nonorthogonal) dic- tionaries via ℓ1 minimization, Proceedings of the National Academy of Sciences, 100 (2003), pp. 2197–2202

    work page 2003

  22. [22]

    M. Du, Y. Chen, and D. Zhang , Discover: Deep identification of symbolically concise open- form partial differential equations via enhanced reinforcement learning, Physical Review Research, 6 (2024), p. 013182. 37

    work page 2024

  23. [23]

    Ducos, A

    A. Ducos, A. Denizot, T. Guyet, and H. Berry , Evaluating PDE discovery methods for multiscale modeling of biological signals , arXiv preprint arXiv:2506.20694, (2025)

    work page arXiv 2025

  24. [24]

    J. Fan, L. Kong, L. Wang, and N. Xiu, Variable selection in sparse regression with quadratic measurements, Statistica Sinica, 28 (2018), pp. 1157–1178

    work page 2018

  25. [25]

    Flandoli, M

    F. Flandoli, M. Gubinelli, and E. Priola , Well-posedness of the transport equation by stochastic perturbation, Invent. Math., 180 (2010), pp. 1–53

    work page 2010

  26. [26]

    Fornberg, Generation of finite difference formulas on arbitrarily spaced grids , Mathematics of computation, 51 (1988), pp

    B. Fornberg, Generation of finite difference formulas on arbitrarily spaced grids , Mathematics of computation, 51 (1988), pp. 699–706

    work page 1988

  27. [27]

    Gerardos and P

    A. Gerardos and P. Ronceray , Principled model selection for stochastic dynamics , arXiv preprint arXiv:2501.10339, (2025)

    work page arXiv 2025

  28. [28]

    W. W. Hager and H. Zhang , A new conjugate gradient method with guaranteed descent and an efficient line search , SIAM Journal on optimization, 16 (2005), pp. 170–192

    work page 2005

  29. [29]

    , A survey of nonlinear conjugate gradient methods, Pacific journal of Optimization, 2 (2006), pp. 35–58

    work page 2006

  30. [30]

    Hairer, Solving the KPZ equation , Ann

    M. Hairer, Solving the KPZ equation , Ann. of Math. (2), 178 (2013), pp. 559–664

    work page 2013

  31. [31]

    R. Y. He, H. Liu, W. Liao, and S. H. Kang, IDENT review: Recent advances in identification of differential equations from noisy data , arXiv preprint arXiv:2506.07604, (2025)

    work page arXiv 2025

  32. [32]

    R. Y. He, H. Liu, and H. Liu , Group projected subspace pursuit for block sparse signal recon- struction: Convergence analysis and applications, Applied and Computational Harmonic Analysis, 75 (2025), p. 101726

    work page 2025

  33. [33]

    He, S.-H

    Y. He, S.-H. Kang, W. Liao, H. Liu, and Y. Liu , Robust identification of differential equa- tions by numerical techniques from a single set of noisy observation , SIAM Journal on Scientific Computing, 44 (2022), pp. A1145–A1175

    work page 2022

  34. [34]

    Y. He, H. Zhao, and Y. Zhong , How much can one learn a partial differential equation from its solution?, Found. Comput. Math., 24 (2024), pp. 1595–1641

    work page 2024

  35. [35]

    D. J. Higham and P. E. Kloeden , An introduction to the numerical simulation of stochastic differential equations, Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA,

    work page

  36. [36]

    Hong and L

    J. Hong and L. Sun , Symplectic integration of stochastic Hamiltonian systems , vol. 2314 of Lecture Notes in Mathematics, Springer, Singapore, [2022] ©2022

    work page 2022

  37. [37]

    S. H. Kang, W. Liao, and Y. Liu, Ident: Identifying differential equations with numerical time evolution, Journal of Scientific Computing, 87 (2021), pp. 1–27

    work page 2021

  38. [38]

    Kunita, Stochastic flows and stochastic differential equations , vol

    H. Kunita, Stochastic flows and stochastic differential equations , vol. 24 of Cambridge Studies in Advanced Mathematics, Cambridge University Press, Cambridge, 1990

    work page 1990

  39. [39]

    Li and J

    Y. Li and J. Duan , A data-driven approach for discovering stochastic dynamical systems with non-Gaussian L´ evy noise, Physica D: Nonlinear Phenomena, 417 (2021), p. 132830

    work page 2021

  40. [40]

    Liu and M

    W. Liu and M. R¨ockner, Stochastic partial differential equations: an introduction, Universitext, Springer, Cham, 2015

    work page 2015

  41. [41]

    Z. Long, Y. Lu, and B. Dong , PDE-Net 2.0: Learning PDEs from data with a numeric- symbolic hybrid deep network , Journal of Computational Physics, 399 (2019), p. 108925

    work page 2019

  42. [42]

    L´opez-Fern´andez, C

    M. L´opez-Fern´andez, C. Palencia, and A. Sch ¨adle, A spectral order method for inverting sectorial Laplace transforms, SIAM J. Numer. Anal., 44 (2006), pp. 1332–1350. 38

    work page 2006

  43. [43]

    Y. C. Mathpati, T. Tripura, R. Nayek, and S. Chakraborty, Discovering stochastic par- tial differential equations from limited data using variational Bayes inference , Computer Methods in Applied Mechanics and Engineering, 418 (2024), p. 116512

    work page 2024

  44. [44]

    D. A. Messenger and D. M. Bortz , Weak SINDy for partial differential equations , Journal of Computational Physics, 443 (2021), p. 110525

    work page 2021

  45. [45]

    T. T. Nguyen, C. Soussen, J. Idier, and E.-H. Djermoune, NP-hardness of ℓ0 minimization problems: revision and extension to the non-negative setting , in 2019 13th International conference on Sampling Theory and Applications (SampTA), IEEE, 2019, pp. 1–4

    work page 2019

  46. [46]

    Pazy, Semigroups of linear operators and applications to partial differential equations , vol

    A. Pazy, Semigroups of linear operators and applications to partial differential equations , vol. 44 of Applied Mathematical Sciences, Springer-Verlag, New York, 1983

    work page 1983

  47. [47]

    J. B. Reade, Eigenvalues of positive definite kernels. II, SIAM J. Math. Anal., 15 (1984), pp. 137– 142

    work page 1984

  48. [48]

    S. H. Rudy, S. L. Brunton, J. L. Proctor, and J. N. Kutz , Data-driven discovery of partial differential equations, Science advances, 3 (2017), p. e1602614

    work page 2017

  49. [49]

    Salinetti and R

    G. Salinetti and R. J.-B. Wets , On the convergence of closed-valued measurable multifunc- tions, Transactions of the American Mathematical Society, 266 (1981), pp. 275–289

    work page 1981

  50. [50]

    W. Song, S. Jiang, G. Camps-Valls, M. Williams, L. Zhang, M. Reichstein, H. Vereecken, L. He, X. Hu, and L. Shi , Towards data-driven discovery of governing equa- tions in geosciences, Communications Earth & Environment, 5 (2024), p. 589

    work page 2024

  51. [51]

    Stephany and C

    R. Stephany and C. Earls , Weak-PDE-LEARN: A weak form based approach to discovering PDEs from noisy, limited data , Journal of Computational Physics, 506 (2024), p. 112950

    work page 2024

  52. [52]

    S. A. Stouffer, E. A. Suchman, L. C. DeVinney, S. A. Star, and R. M. Williams Jr , The american soldier: Adjustment during army life. (studies in social psychology in world war ii), vol. 1, (1949)

    work page 1949

  53. [53]

    C. Tang, R. Y. He, and H. Liu, WG-IDENT: Weak group identification of PDEs with varying coefficients, arXiv preprint arXiv:2504.10212, (2025)

    work page arXiv 2025

  54. [54]

    M. Tang, W. Liao, R. Kuske, and S. H. Kang, Weakident: Weak formulation for identifying differential equation using narrow-fit and trimming, Journal of Computational Physics, 483 (2023), p. 112069

    work page 2023

  55. [55]

    M. Tang, H. Liu, W. Liao, and S. H. Kang , Fourier features for identifying differential equations (fourierident), arXiv preprint arXiv:2311.16608, (2023)

    work page arXiv 2023

  56. [56]

    Tripura and S

    T. Tripura and S. Chakraborty, A sparse Bayesian framework for discovering interpretable nonlinear stochastic dynamical systems with gaussian white noise , Mechanical Systems and Signal Processing, 187 (2023), p. 109939

    work page 2023

  57. [57]

    Tripura, S

    T. Tripura, S. Panda, B. Hazra, and S. Chakraborty , Data-driven discovery of inter- pretable Lagrangian of stochastically excited dynamical systems , Computer Methods in Applied Mechanics and Engineering, 427 (2024), p. 117032

    work page 2024

  58. [58]

    Vogel, A stochastic approach to stability in stochastic programming, Journal of Computational and Applied Mathematics, 56 (1994), pp

    S. Vogel, A stochastic approach to stability in stochastic programming, Journal of Computational and Applied Mathematics, 56 (1994), pp. 65–96

    work page 1994

  59. [59]

    H. Xu, H. Chang, and D. Zhang , DLGA-PDE: Discovery of PDEs with incomplete candidate library via combination of deep learning and genetic algorithm , Journal of Computational Physics, 418 (2020), p. 109584

    work page 2020

  60. [60]

    Zhang and Y

    Z. Zhang and Y. Liu , A robust framework for identification of PDEs from noisy data , Journal of Computational Physics, 446 (2021), p. 110657. 39

    work page 2021